Artificial intelligence across the stone-associated urinary infection pathway: imaging, urosepsis prediction, and antimicrobial stewardship

Complicated urinary tract infection associated with urinary stone disease remains clinically challenging because obstruction, infected stones, delayed microbiological confirmation, and antimicrobial resistance may converge in the same patient. Artificial intelligence can support this pathway when it is framed as clinical risk integration rather than isolated image recognition. We conducted a structured narrative search of PubMed/MEDLINE, Scopus, Web of Science, and Google Scholar from database inception to 31 July 2026, supplemented by reference chaining and relevant clinical and reporting guidelines. The synthesis distinguishes direct evidence in stone-associated infection and perioperative infectious outcomes from indirect evidence in stone detection, volumetry, recurrence, and general UTI stewardship. Radiomics and machine learning studies show promising discrimination, but the evidence remains predominantly retrospective and frequently lacks external validation, complete calibration reporting, and prospective impact assessment. Artificial intelligence should therefore be treated as an adjunct to urological assessment, cultures, imaging, antimicrobial stewardship, and timely source control. For African settings, pragmatic validation must account for local resistance patterns, uneven access to CT and microbiology, referral delays, infrastructure, and the availability of urgent drainage.